A kind of exposure dose determines method and system
Technical field
The present invention relates to medicine radiotherapy fields, determine method and system more particularly to a kind of exposure dose.
Background technique
Radiotherapy and surgical operation therapy, internal medicine chemotherapy and the three big means for becoming oncotherapy.Statistical data shows, about
65% patient needs to use radiotherapy during oncotherapy, part cancer kind (such as early stage nasopharyngeal carcinoma, lung cancer, breast cancer, preceding
Column gland cancer) it can be cured by radiotherapy, radiotherapy is to the contribution degree of tomor rejection up to 40%.
Radiotherapy general steps process is as follows:
The first step, as shown in Figure 1, simulator locating, i.e., carry out the medical image imaging under the conditions of Postural immobilization to patient.
Second step, as shown in Fig. 2, radiotherapy plan, i.e., by the image transmitting of previous step to radiotherapy planning system
(treatmentplanning system, TPS) software is simulated X-ray incident angle, shape, energy on the software, and is led to
The software calculation of radiation dose is crossed, patient's in-vivo tumour and its surrounding jeopardize the exposure dosage point of organ after simulation radiation therapy delivery
Cloth situation carries out statistic quantification assessment by way of dose-volume histogram.
Third step, implementing plan, i.e., above-mentioned radiotherapy planning submit doctor after radiotherapy doctor and radiation physics Shi Gongtong confirmation
It is executed with linear accelerator, as shown in figure 3, before executing, first by identical fixed device, by patient by " first step is simulated
Position when positioning " is retightened in clinac therapeutic bed, i.e. " pendulum position ";Then it is same to enable clinac
Machine image system, the CBCT imaging system that white arrow is directed toward in Fig. 3, in the case where patient is without departing from therapeutic bed, at about 2 points
In the time of clock, fast imaging is carried out to patient's therapentic part;After again, by the image of fast imaging and " second step radiotherapy planning
Patient image used in design " carries out image registration, the position chanP disclosed using registration result, mobile accelerator therapy
Bed, makes patient body position be restored to " first step simulator locating " and " second step radiotherapy plan " identical position, such as Fig. 4 institute
Show;Finally, calling the radiotherapy planning of the patient, implement radiotherapy for patient.
Existing radiotherapy planning method, such as " second step, radiotherapy plan " is described above, existing radiotherapy plan
Process be anatomical map and functional image guidance under formulate, anatomical map (such as CT, MRI) reflect patient's body tumour and
Normal tissue anatomical structure and position, the portion inside patient tumors that functional image (such as f-MRI, PET, SPECT) reflects
Decomposing biological characteristic adjusts the parameter (including beam shape, intensity) of irradiation field as reference in plan design process, from
And simulate and calculate patient's body dosage distribution situation, complete plan design.
If " second step, radiotherapy plan " is described above, the formulation process of existing radiotherapy planning has references to dissection
Image (such as CT, MRI) and functional image (such as f-MRI, PET, SPECT), limitation are embodied in:
Guided with anatomical map and formulate radiotherapy planning, anatomical map can only reflect tumour and normal surrounding tissue position and
Structure, but can not reflect biological function information (glucose metabolism of such as inside tumor is horizontal, weary oxygen level, angiogenesis situation,
There is size of complication risk etc. in normal tissue);And the biological nature of people's in-vivo tumour or normal tissue, on the one hand very
Determined in big degree radiotherapy success or failure (such as the weary oxygen region for inside tumor, the presence in weary oxygen region and tumour it is remote
End transfer and poor prognosis have directly related property, and numerous studies confirm that the tumour cell in weary oxygen region resists radioactive ray,
If not implementing high-dose irradiation, tumour possibly can not be killed effectively;For another example shown in Fig. 5, although white arrow institute
Show that tumor region does not occur the diminution in volume from CT anatomical structure, but CT perfusion functional image shows that blood flow lowers,
That is tumor promotion cell has substantially reduced, and actual therapeutic is effective;For another example, for easily there is the normal of radiation pneumonitis
Function lung is not protected targetedly, then patient receives the radiation pneumonitis that lethal is likely to occur after radiotherapy), it is another
Aspect individuation difference is huge, and there may be huge biological nature difference (for example same to put by two patients of same class tumour
After treating exposure dose implementation, radiosensible patient tumors subside obviously, and the patient tumors of Radioresistence are constant or even increase
Grow), therefore, it is very crucial and necessary for capable of reflecting that the factor of biological function information participates in the formulation process of radiotherapy planning
's.
It is guided with functional image and formulates radiotherapy planning, although reflecting specific biological function information (such as inside tumor
There is size of complication risk etc. in glucose metabolism situation, weary oxygen situation, angiogenesis situation, normal tissue), but it exists
Many limitations, including:
1. position and anatomic information can not be accurately reflected, f-MRI, PET, SPECT image are in reflection partial region biology function
While energy, the exact anatomic location in the region can not be accurately reflected, it can only be by carrying out image co-registration (i.e. two with anatomical map
The superposition of kind image), functional area could be directed toward to corresponding anatomical position, as Figure 6-Figure 8, the highlight regions in Fig. 8
For the region with particular organisms function, such as tumour glucose hypermetabolism or tumor hypoxia region;But this " direction " is again past
Toward deficient accurate, the Image Fusion because its precision places one's entire reliance upon, and often there is deformation, scale not in the image of different modalities
The problems such as unified, it can not accomplish one-to-one correspondence truly.
2. functional image somewhat expensive, few then more than 1,000 yuan/time (SPECT), more then nearly ten thousand yuan/time (PET, non-medical insurance reimbursements
Project, completely at one's own expense), implementation difficulty is big in most of patients, can not benefit more patients.
3. the functional images such as PET, SPECT need to inject radioactive isotope in patient's body, additional spoke is brought for patient
Penetrate, thus also limit same patient receive in treatment course functional image scanning total degree should not it is excessive (usually only in
Radiotherapy receives the scanning of functional image before starting, sweep again at most after treatment primary, i.e. front and back receives 2 subfunction images altogether and sweeps
It retouches).
4. a kind of functional image can only disclose a kind of biological function characteristic, as 18F-FETNIM PET can disclose inside tumor
Weary oxygen region, prompt preferably to increase exposure dose to this region during treatment plan, but 18F-FETNIM PET also can only be anti-
Reflect " tumor hypoxia " this feature, the other biologicals functional character such as no method interpretation inside tumor such as angiogenesis, glycometabolism, mesh
It is preceding still to accomplish to inject after human body without a kind of tracer while show multiple functions characteristic.
In view of the above problems, a kind of method identification by local image group feature extraction of foreign study which
In be tumour, where be non-tumour, for being identified as the region of tumour, higher exposure dose is given in Patients During Radiotherapy, and
Then give relatively low prophylactic irradiation dosage in region for being identified as non-tumour.But due to the life of different zones different location
Object feature is different, if giving different exposure doses with tumor region and non-tumor region, due to the life inside tumor region
Tumour cell then can not all be killed in tumor region with identical exposure dose, can not be directed to tumour by object feature difference
The tumour of different biological features gives exposure dose in region, influences the curative effect of radiotherapy.
Summary of the invention
The object of the present invention is to provide a kind of exposure doses to determine method and system, to solve not being directed in the prior art
The problem of tumour of different biological features gives exposure dose, influences radiotherapy effect in tumor region.
To achieve the above object, the present invention provides following schemes:
A kind of exposure dose determines method, including:
Utilize the retrospective pixel for marking and determining and there is in area-of-interest biological characteristic in Radiation Therapy Simulation positioning image;
The Radiation Therapy Simulation positioning image includes computerized tomography image, nuclear magnetic resonance image and positron emission fault image;Institute
State glucose metabolism situation, weary oxygen region, oxygen-rich area that biological characteristic includes inside tumor, angiogenesis situation and normal
There is the size of complication risk in tissue;
Local image group feature is extracted according to the pixel with biological characteristic;The local image group feature
Including grey level histogram intensity, tumor shape feature, textural characteristics, Gaussian Laplce filtering characteristics and wavelet character;
Obtain local image group feature to be measured;
The positive region of the local image group feature to be measured is identified according to the local image group feature;
Three-dimensional reconstruction is carried out to the peripheral boundary of the positive region, determines 3-D image;The 3-D image is display
The 3-D image of the biological characteristic;
The exposure dose of different zones different location is determined according to the 3-D image.
Optionally, the pixel according to biological characteristic extracts local image group feature, specifically includes:
Using method pixel-by-pixel to the pixel with biological characteristic point pixel-by-pixel, using grey level histogram feature extraction
Method, texture feature extraction method, Gaussian Laplce's filtering characteristics extraction method and wavelet decomposition feature extraction extract part
Image group feature.
Optionally, the positive that the local image group feature to be measured is identified according to the local image group feature
Region specifically includes:
The local image group feature is screened, determines optimal feature subset;
The machine learning model for having supervision is established according to the optimal feature subset;
The positive region of the local image group feature to be measured is identified according to the machine learning model for having supervision.
Optionally, described that the local image group feature is screened, it determines optimal feature subset, specifically includes:
The local image group feature is screened by Method for Feature Selection, determines optimal feature subset;The spy
Levying back-and-forth method includes maximal correlation minimal redundancy method.
Optionally, the exposure dose that different zones different location is determined according to the 3-D image, specifically includes:
Exposure dose parameter is determined according to the 3-D image;The radiation parameters include incident angle, intensity and shape
Shape;
The exposure dose of different zones different location is determined according to the exposure dose parameter.
A kind of exposure dose determines system, including:
Pixel determining module, for determining the interior tool of area-of-interest in Radiation Therapy Simulation positioning image using retrospective label
There is the pixel of biological characteristic;Radiation Therapy Simulation positioning image includes computerized tomography image, nuclear magnetic resonance image and just
Emission tomography image;The biological characteristic include the glucose metabolism situation of inside tumor, weary oxygen region, oxygen-rich area,
There is the size of complication risk in angiogenesis situation and normal tissue;
Local image group characteristic extracting module, for extracting local image according to the pixel with biological characteristic
Group learns feature;The local image group feature includes grey level histogram intensity, tumor shape feature, textural characteristics, Gaussian
Laplce's filtering characteristics and wavelet character;
Local image group feature to be measured obtains module, for obtaining local image group feature to be measured;
Positive region identification module, for identifying the local image group to be measured according to the local image group feature
The positive region of feature;
Three-dimensional reconstruction module carries out three-dimensional reconstruction for the peripheral boundary to the positive region, determines 3-D image;Institute
Stating 3-D image is the 3-D image for showing the biological characteristic;
Exposure dose determining module, for determining the exposure dose of different zones different location according to the 3-D image.
Optionally, the local image group characteristic extracting module specifically includes:
Local image group feature extraction unit, for using method pixel-by-pixel to the pixel with biological characteristic by
Pixel, use grey level histogram feature extraction, texture feature extraction method, Gaussian Laplce's filtering characteristics extraction method with
And wavelet decomposition feature extraction extracts local image group feature.
Optionally, the positive region identification module specifically includes:
Screening unit determines optimal feature subset for screening to the local image group feature;
Model building module, for establishing the machine learning model for having supervision according to the optimal feature subset;
Positive region recognition unit, for there is the machine learning model of supervision to identify the local image to be measured according to
Group learns the positive region of feature.
Optionally, the screening unit specifically includes:
Screening subelement determines optimal for being screened by Method for Feature Selection to the local image group feature
Character subset;The Method for Feature Selection includes maximal correlation minimal redundancy method.
Optionally, the exposure dose determining module specifically includes:
Exposure dose parameter determination unit, for determining exposure dose parameter according to the 3-D image;The irradiation ginseng
Number includes incident angle, intensity and shape;
Exposure dose determination unit, for determining the irradiation agent of different zones different location according to the exposure dose parameter
Amount.
The specific embodiment provided according to the present invention, the invention discloses following technical effects:The present invention provides one kind
Exposure dose determines method and system, by determining the pixel of biological characteristic, extracts local image group according to the pixel
Feature is determined the positive region of local image group feature to be measured by the local image group feature, and then is established and can be shown
The 3-D image of the biological characteristic of different zones different location is shown, to determine different zones different location according to 3-D image
Exposure dose also can be for different swollen according to the biological characteristic difference inside tumor region even if being all tumor region
Tumor gives different exposure doses, so as to carry out giving exposure dose according to the actual conditions of tumour, to improve radiotherapy
Curative effect.
Detailed description of the invention
It in order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, below will be to institute in embodiment
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention
Example, for those of ordinary skill in the art, without any creative labor, can also be according to these attached drawings
Obtain other attached drawings.
Fig. 1 is simulator locating flow chart provided by the present invention;
Fig. 2 is formulation radiotherapy planning flow chart provided by the present invention
Practical figure when Fig. 3 is progress simulator locating provided by the present invention;
Fig. 4 is the practical figure provided by the present invention when carrying out formulation radiotherapy planning;
Fig. 5 is that schematic diagram is perfused in CT provided by the present invention;
Fig. 6 is computer tomography figure provided by the present invention;
Fig. 7 is positron emission tomography figure provided by the present invention;
Fig. 8 is computer tomography figure provided by the present invention and the fused figure of positron emission tomography figure
As schematic diagram;
Fig. 9 is that exposure dose provided by the present invention determines method flow diagram;
Figure 10 is that positive region provided by the present invention marks schematic diagram;
Figure 11 is local image group feature extraction flow chart provided by the present invention;
Figure 12 is optimal feature subset selection flow chart provided by the present invention;
Figure 13 is the machine learning identification model Establishing process figure provided by the present invention for having supervision;
Figure 14 is the positive region identification process figure of local image group feature to be measured provided by the present invention;
Figure 15 is that exposure dose provided by the present invention determines system construction drawing.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
The object of the present invention is to provide a kind of exposure doses to determine method and system, can according to tumour actual conditions into
Row gives exposure dose, improves radiotherapy effect.
In order to make the foregoing objectives, features and advantages of the present invention clearer and more comprehensible, with reference to the accompanying drawing and specific real
Applying mode, the present invention is described in further detail.
Image group (Radiomics is also translated into " radiation group ", but the saying of " image group " is more popular at home,
State natural sciences fund committee discipline classification in, just had recently a new three-level subject be called " image group and
Artificial intelligence " therefore continues to use the saying of " image group " herein) be an emerging diagnosing tumor and auxiliary detection technique, especially
It is rapidly developed in nearest decades.Image group refers to will using the data characteristics extraction algorithm largely automated
The image data of region of interest is converted into single order or high level data, by data mining and analyzes its profound relationship, goes forward side by side
The predictive value of raising the clinical diagnosis accuracy and prognosis of one step.
There is research to confirm, the local image group characteristic index extracted from the images such as CT, MRI, PET and tumour itself
Biological characteristics and clinical treatment curative effect have significant correlation, such as:1. the certain local image groups extracted from tumor region
It is related with inside tumor heterogeneity to learn characteristic index, is directed toward different zones glucose metabolism situation in tumour, is weary oxygen or richness
Oxygen, if having angiogenesis;2. certain local image group characteristic indexs and the tumour extracted from tumor region receive radiotherapy or
Reaction susceptibility after chemotherapy is related, these indexs can be used for predicting chemicotherapy curative effect;3. from the certain of lung tissue extracted region
Local image group characteristic index, it is related to occur a possibility that radiation pneumonitis after receiving radiotherapy with patient, can be used for treating
There is the risk of complication after receiving radiotherapy in preceding prediction patient;4. being mentioned out of oropharyngeal cancer CT picture of patient is reflected tumor region
There is far-end transfer there are correlation in local image group feature and the patient taken, and it is remote to can be used for predicting that patient occurs before the treatment
Hold the risk of transfer.
The studies above is all published in the top periodical of industry field, is widely recognized as in the industry.
But it is existing research all concentrate on disclosing it is related between different characteristic index and different clinical manifestations
Property, and rarely have research and inquirement to be used to these indexs guide formulation radiotherapy planning.It is related research shows that in MRI image extract before
The part local image group characteristic index of adenocarcinoma patients' prostate region is arranged to identify the growth of tumour cell in prostate
These regions are added on corresponding CT image then by image registration and regard as tumor target, then at this by region
Radiotherapy planning is formulated to the tumor target on CT image;But equally it is tumor region, the weary oxygen region in tumour is needed than rich
The tumour in oxygen region, which obtains higher exposure dose, to kill cell, due to different zones different location biological characteristic not
With, therefore, if giving identical irradiation with identical tumour in the radiotherapy planning of the prior art different tumour or different zones
Dosage, radiotherapeutic effect are very poor.
Fig. 9 is that exposure dose provided by the present invention determines method flow diagram, as shown in figure 9, a kind of exposure dose determines
Method, including:
Step 901:Determine in Radiation Therapy Simulation positioning image that there is biological characteristic in area-of-interest using retrospective label
Pixel;The Radiation Therapy Simulation positioning image includes that computerized tomography image, nuclear magnetic resonance image and positron emission are disconnected
Layer image;The biological characteristic includes the glucose metabolism situation of inside tumor, weary oxygen region, oxygen-rich area, angiogenesis feelings
There is the size of complication risk in condition and normal tissue.
As shown in Figure 10, retrospective label enters the Radiation Therapy Simulation positioning image of the patient of training set and verifying collection (as counted
Calculation machine tomographic imaging, Magnetic resonance imaging, positron emission tomography, conical beam CT, Single Photon Emission Computed tomography
Imaging, megavolt grade CT, electronic portal imaging device EPID and barivm meal fluoroscopy (screem) image etc.) in there is biological characteristic sun in area-of-interest
Property pixel, i.e., by retrospective analysis, by the pathological section image of patient, functional image (such as 18F-FETNIM PET/CT
Can non-invasively show the weary oxygen region of inside tumor) or other reflection region of interest biological characteristics (such as lung tissue generation lung radiation injuries
Scorching region) image and Radiation Therapy Simulation position Image registration, then according to pathological section image, functional image or other reflections
Whether the pixel (pixels, P) in the image of region of interest biological characteristic has certain biological characteristic interested, to Radiation Therapy Simulation
(positive pixel label is marked in corresponding pixel in positioning image region of interest (region ofinterest, ROI)
1, negative pixel label 0).
Wherein, define in a certain region of interest ROI all pixels point be set P=plable 1, plabel2 ...,
Plabeln }, wherein n indicates that pixel quantity, lable indicate the label of pixel (1 indicates positive, and 0 indicates negative).
Step 902:Local image group feature is extracted according to the pixel with biological characteristic;The local image
It includes grey level histogram intensity, tumor shape feature, textural characteristics, Gaussian Laplce filtering characteristics and small that group, which learns feature,
Wave characteristic.
As shown in figure 11, extract it is positive in each region of interest ROI (such as tumor target) in Radiation Therapy Simulation positioning image and
The local image group feature of negative pixel.
1. local image group feature extraction algorithm includes but is not limited to
Grey level histogram feature extraction algorithm, the feature quantity extracted are designated as N1;
Texture Segmentation Algorithm, the feature quantity extracted are designated as N2;
Gaussian Laplce's filtering characteristics extraction algorithm, the feature quantity extracted are designated as N3;
Wavelet decomposition feature extraction algorithm, the feature quantity extracted are designated as N4;
Other feature extraction extraction algorithms, the feature quantity extracted are designated as N5;
Therefore general characteristic number N=N1+N2+N3+N4+N5;Characteristic value fj (p) expression, p ∈ P, j ∈ { 1 ..., N }).
2. the local shape factor based on pixel.Traverse pixel all in every ROI in Radiation Therapy Simulation positioning image
Point calculates surrounding δ × δ × δ (being directed to 3-D image) or δ × δ (being directed to two dimensional image) neighborhood window centered on each pixel
The local image group characteristic value of mouth, wherein δ is the odd number more than or equal to 3, and the calculating for local feature, the value of δ is each time
3,5,7,9 are taken respectively, are chosen in subsequent modeling and verification step so that the optimal δ value of accuracy of identification.For borderline
Pixel, using symmetrical filling, the value of filler pixels is the mirror reflection of the boundary pixel.
3. by step 2. based on the local shape factor of pixel, under specific δ value, for each pixel (p ∈
P), the feature vector Flabel i={ f of a 1 × N-dimensional can be obtained1(plabeli), f2(plabel i) ..., fj(plabel
I) }, j ∈ { 1 ..., N }, i ∈ { 1 .., n }, label ∈ { 1,0 } (that is, for pixel p label i, can extract N number of feature
Value, respectively f1(pi), f2(pi) ..., fj(pi), it is denoted as vector Flabel i).Therefore, for all pixels in ROI, feature
Collection is combined intolabel∈{1,0}.Each pixel (label=1/0) is used as one
Sample, the sample have N-dimensional local image group feature.
Step 903:Obtain local image group feature to be measured.
Step 904:The hot spot of the local image group feature to be measured is identified according to the local image group feature
Domain.
The step 904 specifically includes:
As shown in figure 12, the local image group feature is screened, determines optimal feature subset.Based on training
Collection, passes through relevant feature selecting algorithm (including but not limited to maximal correlation minimal redundancy method (Minimum Redundancy-
Maximum Relevance, mRMR)), filter out with tag along sort between maximum correlation, feature have minimal redundancy
The characteristic parameter of degree, the foundation for next step prediction model.By the step, optimal feature subset G={ f is obtained1(p),f2
(p),…,fk(p)}.Each there are the pixel (label=1/0) of label and its local image group feature to join as a sample
With feature selecting, positive and negative pixel both participates in the process, and selected optimal feature subset is the same set.
As shown in figure 13, the machine learning model for having supervision is established according to the optimal feature subset.1. modeling:Based on instruction
Practice the machine that the input of image group feature described in the optimal characteristics set G for the pixel (label=1/0) for having label is had supervision by collection
Device learning algorithm, for establishing the machine learning identification model for having supervision.Each there is pixel (label=1/0) conduct of label
One sample participates in the foundation of model.
2. verifying:Based on the accuracy of identification of verifying collection data verification model, and select so that accuracy of identification is optimal
Feature extraction window (δ opt).
As shown in figure 14, the local image group feature to be measured is identified according to the machine learning model for having supervision
Positive region.For the new positioning image of local image group feature to be formulated guidance radiotherapy planning, with δoptThe neighborhood of size
Window extracts local image group feature described in set G pixel-by-pixel;The characteristic parameter of each pixel is inputted the 4th step to establish
Identification model, the biological function expression label which will export current pixel is (positive:1, feminine gender is 0).By same biological function
Pixel that can be positive is designated as uniform color, i.e., shows biological function positive region on each tension fault image.
Step 905:Three-dimensional reconstruction is carried out to the peripheral boundary of the positive region, determines 3-D image;The three-dimensional figure
As the 3-D image to show the biological characteristic.
The positive region peripheral boundary that the step 904 is identified carries out three-dimensional reconstruction, in planning system image space
In reconstruct the stereochemical structure in the biological function area, determine 3-D image.
Step 906:The exposure dose of different zones different location is determined according to the 3-D image.
Three-dimensional image reconstruction go out biological characteristic on the basis of, be arranged irradiation field parameter, including incident angle, intensity,
Shape, and dose-volume or biotic factor such as Normal Tissue Complication probability, tumor control probability are given about for the region
Beam condition, and then carry out plan design and objective optimization.
The present invention is by the way that from medical image, (medical image includes computer tomography, Magnetic resonance imaging, positive electricity
Sub- emission tomography, conical beam CT, single photon emission computerized tomography, megavolt grade CT, electronic portal imaging device EPID
And barivm meal fluoroscopy (screem) image, but these images are not limited to, future may have other images to formulate for radiotherapy planning, be suitable for
The technology) in the textural characteristics of body local anatomic region that extract, human body topographic region can be tumor region,
It may be normal tissue regions, prediction patient receives the variation being likely to occur after radiotherapy, and (variation including tumour is as subsided, resisting
Constant, increase, the complication being likely to occur including normal tissue), to targetedly formulate and put for patient as guidance
Plan is treated, to realize targeted individuation accurate radiotherapy, patient is promoted and treats benefit.
Figure 15 is that exposure dose provided by the present invention determines system construction drawing, and as shown in figure 15, a kind of exposure dose is true
Determine system, including:
Pixel determining module 1501, for determining area-of-interest in Radiation Therapy Simulation positioning image using retrospective label
The interior pixel with biological characteristic;Radiation Therapy Simulation positioning image include computerized tomography image, nuclear magnetic resonance image with
And positron emission fault image;The biological characteristic includes the glucose metabolism situation of inside tumor, weary oxygen region, oxygen-rich area
There is the size of complication risk in domain, angiogenesis situation and normal tissue.
Local image group characteristic extracting module 1502, for extracting part according to the pixel with biological characteristic
Image group feature;The local image group feature includes grey level histogram intensity, tumor shape feature, textural characteristics, height
This type Laplce filtering characteristics and wavelet character.
The local image group characteristic extracting module 1502 specifically includes:Local image group feature extraction unit is used
In using method pixel-by-pixel to the pixel with biological characteristic point pixel-by-pixel, using grey level histogram feature extraction, line
It manages feature extraction, Gaussian Laplce's filtering characteristics extraction method and wavelet decomposition feature extraction and extracts local image group
Learn feature.
Local image group feature to be measured obtains module 1503, for obtaining local image group feature to be measured.
Positive region identification module 1504, for identifying the local image to be measured according to the local image group feature
Group learns the positive region of feature.
The positive region identification module 1504 specifically includes:
Screening unit determines optimal feature subset for screening to the local image group feature;Model foundation
Module, for establishing the machine learning model for having supervision according to the optimal feature subset;Positive region recognition unit is used for root
The positive region of the local image group feature to be measured is identified according to the machine learning model for having supervision.
The screening unit specifically includes:Subelement is screened, for passing through Method for Feature Selection to the local image group
Feature is screened, and determines optimal feature subset;The Method for Feature Selection includes maximal correlation minimal redundancy method.
Three-dimensional reconstruction module 1505 carries out three-dimensional reconstruction for the peripheral boundary to the positive region, determines three-dimensional figure
Picture;The 3-D image is the 3-D image for showing the biological characteristic.
Exposure dose determining module 1506, for determining the irradiation agent of different zones different location according to the 3-D image
Amount.
The exposure dose determining module 1506 specifically includes:Exposure dose parameter determination unit, for according to described three
Dimension image determines exposure dose parameter;The radiation parameters include incident angle, intensity and shape;Exposure dose determines single
Member, for determining the exposure dose of different zones different location according to the exposure dose parameter.
It is existing research only by local image group Feature Extraction Technology prostate region identification where be tumour, where
In be non-tumour, for being identified as the region of tumour, give higher exposure dose in plan design process, and for identification
Relatively low prophylactic irradiation dosage is then given for the region of non-tumour.
And the present invention provides a kind of determination method and system of exposure dose, on the one hand, can not only recognise that
For tumor region, additionally it is possible to identification is biological variability inside tumor region, such as:It is equally tumor region, in tumour
Weary oxygen region, which needs to obtain higher exposure dose than the tumour of oxygen-rich area, to kill cell, and therefore, the present invention is base
Method and system are determined in the exposure dose of comprehensive identification tumour to local image group feature;On the other hand, the present invention is not
It can only identify tumour, additionally it is possible to identify the biological property of normal tissue regions, such as:Plan is carried out for breast tumor to set
Before meter, the feature extraction of local image group just is carried out to lung tissue, in the case where human eye can't see, it can be found that lung tissue
The micro-variations of structure, prompt in time doctor and physics teacher the lung tissue for planning targetedly to reduce certain region when design by
Exposure dose, so that the method for planning of really " cutting the garment according to the figure " is realized, to improve the curative effect of radiotherapy.
Each embodiment in this specification is described in a progressive manner, the highlights of each of the examples are with other
The difference of embodiment, the same or similar parts in each embodiment may refer to each other.For system disclosed in embodiment
For, since it is corresponded to the methods disclosed in the examples, so being described relatively simple, related place is said referring to method part
It is bright.
Used herein a specific example illustrates the principle and implementation of the invention, and above embodiments are said
It is bright to be merely used to help understand method and its core concept of the invention;At the same time, for those skilled in the art, foundation
Thought of the invention, there will be changes in the specific implementation manner and application range.In conclusion the content of the present specification is not
It is interpreted as limitation of the present invention.